Unsupervised Variable Selection: If You Don’t Know, I’m Certainly Not Going to Tell You
From Dan Kelly: "Typically, feature selection is considered a supervised task: you want the inputs that predict the behavior or amount of interest. If you have multiple targets of interest, or if you have wide data, or if you have tall data, or if you have to explain your models to human beings at some point, it may be worthwhile to prescreen candidate inputs with an unsupervised approach. This talk will address some typical approaches and their respective benefits and shortcomings. It does not currently feature any dirty limericks, but if I can come up with a good rhyme for “eigenvalue” by Tuesday, I’ll see what I can do."
Please join us on Tuesday, July 16th for this talk on how to get the most out of your data.
Doors open at 6:30. Talk begins at 7:00.----